Printed out Johann’s recirculating-ball carriage prototype today. Had trouble getting the right support with Slic3r, so I ended manually adding some support in OpenSCAD. The carriages are not as smooth as real metal ones (of course!) but are solid and usable. Can’t beat the price either, as the balls $3 for 100 and each carriage uses 38 of them. This is up there for me with stuff like Emmett’s gear bearings as a shining example of the kind of thing that would be practicaly impossible to do in one piece without 3d printing. Very cool.
A few downsides, though:
They’re a bit noisy, and most of the noise seems to come from the top ball “falling” down. I left one ball out, because with it in the motion got rough, presumably because it forced each ball against the plastic track and probably left no space for imperfections. A tiny increase in racetrack length (or perfect printing) might fix this. It’s easy to find the imperfections (mostly layer seams) but not so easy to fix them, as the inside is not easily accessible.
Another issue is that the assembly fully wraps around the OpenBeam. Johann got things perfect as far as size, though; there’s really no need for the screw to pull the sides together, and you could imagine alternate ways to pull them together, like an eccentric spacer.
Note the failed first print in the pics; I didn’t have a bottom plate to keep the little supports together, and when the first perimeter came on top, they all pulled off the build plate. Easily fixed with a bottom raft to hold the supports.
Printed on a mostly stock Kossel Air with the “real” bal rails and a non-geared extruder pushing 1.75 ultimachine red PLA @ 205C.
Source:
Pics from Johann’s original:
Brandon, did you see my post about scaling them up for 2020?
https://plus.google.com/+StephenKongsle1/posts/3goQFLxu9Z9
In my version I addressed a few of the issues you mention. For example, the adjustment screw in mine doesn’t go on the other sided of extrusion.
I also changed the length of the track so there was enough room for the balls to adjust to some imperfection, but not so much that they clack when hitting each other or the aluminum. That said, the carriages are still too noisy to be used on a machine that would sit on the desk where I work. A machine in the garage, fine. 
I’ve seen a lot of industrial ones use teflon spacers between steel ball bearings to prevent clacking and improve performance via reduced friction and lower surface speeds.
+Stephen, just went back and re-read the thread. I’m using OpenBeam, so your part won’t fit, but I do like the approach to open up one of the sides. Did the resize to have a one ball gap help reduce the noise at all? I think it’s where the vast majority of noise comes from on mine.
@Jerry_Rodberg are the spacers spherical? Trying to visualize them.
@Brandon_Heller The first version I made was a direct scale of Johann’s, and I found, like you, that removing one ball made things run smoother. But the clacking was annoying. So in later versions I adjusted the length such that there’s a little wiggle room for the balls. I can’t remember how much, maybe two mm (overall, not per ball). I think they run as smooth as the original with one ball missing, but without the clacking. Sadly, though, as I said, they’re still noisy.
Maybe add some heavy bearing grease to silence them?
The spacers I remember are basically back to back cups that go around the ball partially. they may slide on the inner race but are designed to not touch the actual bearing surface. I’ll see if I can find a reference, it was only in the really big trucks they started showing up, or in standard rotary bearings.
I’m not aware of anyone who has done longevity tests. Johann mentioned that they wore in a bit for him on his original Deltabot mailing list post. I’m tempted to make another set but with a tiny bit larger race length and a tighter tolerance to the ball. That should reduce the two sources of clacking that I’m aware of.
